java.lang.Object
- gov.nih.mipav.view.renderer.WildMagic.AAM.CDOptimizeBase

Direct Known Subclasses:

CDOptimizeBFGS, CDOptimizeCG, CDOptimizePS, CDOptimizeSA, CDOptimizeSD
```
public abstract class CDOptimizeBase
extends java.lang.Object
```
This is the Java modified version of C++ active appearance model API (AAM_API). It is modified with a subset of required functions for automatic MRI prostate segmentation. AAM-API LICENSE - file: license.txt This software is freely available for non-commercial use such as research and education. Please see the full disclaimer below. All publications describing work using this software should cite the reference given below. Copyright (c) 2000-2003 Mikkel B. Stegmann, mbs@imm.dtu.dk IMM, Informatics & Mathematical Modelling DTU, Technical University of Denmark Richard Petersens Plads, Building 321 DK-2800 Lyngby, Denmark http://www.imm.dtu.dk/~aam/ REFERENCES Please use the reference below, when writing articles, reports etc. where the AAM-API has been used. A draft version the article is available from the homepage. I will be happy to receive pre- or reprints of such articles. /Mikkel ------------- M. B. Stegmann, B. K. Ersboll, R. Larsen, "FAME -- A Flexible Appearance Modelling Environment", IEEE Transactions on Medical Imaging, IEEE, 2003 (to appear) ------------- 3RD PART SOFTWARE The software is partly based on the following libraries: - The Microsoft(tm) Vision Software Developers Kit, VisSDK - LAPACK DISCLAIMER This software is provided 'as-is', without any express or implied warranty. In no event will the author be held liable for any damages arising from the use of this software. Permission is granted to anyone to use this software for any non-commercial purpose, and to alter it, subject to the following restrictions: 1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. 2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software. 3. This notice may not be removed or altered from any source distribution. -- No guarantees of performance accompany this software, nor is any responsibility assumed on the part of the author or IMM. This software is provided by Mikkel B. Stegmann and IMM ``as is'' and any express or implied warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed. In no event shall IMM or Mikkel B. Stegmann be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including, but not limited to, procurement of substitute goods or services; loss of use, data, or profits; or business interruption) however caused and on any theory of liability, whether in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of the use of this software, even if advised of the possibility of such damage. $Revision: 1.4 $ $Date: 2003/04/23 14:49:15 $ Optimization base class. This abstract class is the base class for all classes implementing specific optimization procedures. author: Rune Fisker, 26/1-1999

Author:

Ruida Cheng

Field Summary

Fields
Modifier and Type	Field	Description
`static int`	`CentralDifference`
`int`	`ENumGrad`
`static int`	`eoptBFGS`
`static int`	`eoptConjugateGradient`
`int`	`EOptMethod`
`static int`	`eoptPatternSearch`
`static int`	`eoptSimulatedAnnealing`
`static int`	`eoptSteepestDescent`
`static int`	`eoptUnknown`
`int`	`ETermCode`
`static int`	`etermConsecMaxStepMax`
`static int`	`etermDeltaFuncVal`
`static int`	`etermGradTol`
`static int`	`etermLineSearch`
`static int`	`etermMaxFuncEval`
`static int`	`etermMaxIterations`
`static int`	`etermNoStop`
`static int`	`etermStepTol`
`static int`	`etermUnknown`
`static int`	`FitLine`
`static int`	`ForwardDifference`
`double`	`m_dDeltaFuncVal`	value for stop criterion: abs(f - fplus) < m_dDeltaFuncVal
`double`	`m_dEta`	machine precision.
`double`	`m_dGradTol`	A positive scalar giving the tolerance at which the scaled gradient in considered close enough to zero to terminate the algorithm
`double`	`m_dMachEps`	Machine precision
`double`	`m_dMaxStep`	A positive scalar giving the maximum allowable scaled steplength at any iteration. maxstep is used to prevent steps that would cause the optimazation algorithm to overflow or leave the domain of interest, as well as to detect divergence.
`double`	`m_dStepTol`	A positive scalar giving the tolerance at which the scaled distance between two successive iterated is considered close enough to zero to terminate the algorithm
`double`	`m_dTypF`	positive scalar estimating the magnitude of f(x) near he minimizer x-star.
`int`	`m_eNumGrad`	methods for estimating the gradient nummericaly
`boolean`	`m_fAnalyticGrad`	flag indicating if the gradient is to be calc. analyticaly or nummericaly
`boolean`	`m_fLogFuncValues`	Logical variable which determines whether function parameters and corresponding return values should be stored for later analysis.
`int`	`m_iStopCriteria`	holds the stop criteria in use, e.g m_iStopCriteria = etermMaxFuncEval \| etermMaxIterations
`int`	`m_nConsecMax`	Number of conseccutive past steps whose scaled length was equal to maxstep
`int`	`m_nConsecMaxStepMax`	max number of conseccutive past steps whose scaled length was equal to maxstep to terminate
`int`	`m_nFDigits`	A positive integer specifying the number of reliable digits returned by the objective function FN. fdigits is used to set the parameter n (eta) that is used in the code to specify the relative noise in f(x); the main use of eta is in calculation finite difference step size. eta is set to macheps if fdigits = -1.
`int`	`m_nFuncEval`	counter to the number of function evaluations.
`int`	`m_nGradEval`	counter to the number of gradient evaluations.
`int`	`m_nIterations`	counter to the number of iterations.
`private int`	`m_nMaxFuncEval`	limit for number of function evaluations.
`private int`	`m_nMaxIterations`	A positive integer specifying the maximum number of iterations that may be performed before the algorithm is halted.
`private CDOptimizeFuncBase`	`m_pFuncEvalBase`	pointer to the function to be minimized.
`CDVector`	`m_vFuncVal`
`private CDVector`	`m_vMethodPar`	special parameters used by each optimization method e.g. step size used for calc. nummerical gradient
`CDVector`	`m_vNFuncEval`
`java.util.Vector<CDVector>`	`m_vvFuncParm`

Constructor Summary

Constructors
Constructor Description

CDOptimizeBase()
Constructor

Method Summary

All Methods Instance Methods Abstract Methods Concrete Methods
Modifier and Type	Method	Description
`double`	`EvalFunction(CDVector x)`	function and gradient evaluation methods
`void`	`EvalGradient(CDVector x, CDVector gc, double dFuncVal)`	function and gradient evaluation methods
`int`	`ExactLineSearch(CDVector xc, double fc, CDVector g, CDVector p, CDVector xplus, double[] fplus, boolean[] maxtaken)`	Perform exact line search Input: xc: parameter fc: function value at xc g: gradient at xc p: search direction Output: xplus: new parameter fplus: function value for new parameter maxtaken: max taken in line search return termination code
`CDOptimizeFuncBase`	`GetFuncEvalBase()`	get point to func eval. base.
`int`	`LineSearch(CDVector xc, double fc, CDVector g, CDVector p, CDVector xplus, double[] fplus, boolean[] maxtaken)`	Warpper for line search
`int`	`LineSearch(CDVector xc, double fc, CDVector g, CDVector p, CDVector xplus, double[] fplus, boolean[] maxtaken, boolean fSoft)`
`int`	`MaxFuncEval()`	get the number of evaluations
`int`	`MaxIterations()`	get limit for the number of iterations
`CDVector`	`MethodPar()`	get numerical gradient function
`abstract int`	`Minimize(CDVector x, CDOptimizeFuncBase pFuncEvalBase)`	the Minimize function using analytic gradient
`int`	`MinimizeNum(CDVector x, CDOptimizeFuncBase pFuncEvalBase, CDVector vMethodPar)`
`java.lang.String`	`Name()`	name of optimization methode
`void`	`NumGrad(CDVector x, CDVector gradient, double dFuncVal)`	calculate nummerical gradient function usingdecided gradient calculation method input: x: parameter dFuncVal: function value in x output: gradient: gradient
`int`	`OptMethod()`	name of optimization method
`void`	`SetFuncEvalBase(CDOptimizeFuncBase pFuncEval)`	set point to func eval. base.
`void`	`SetMachineEps()`	----------------------------[ MachineEps ]---------------------------- Calculate machine epsilon Algorithm A1.3.1 - p. 303 Dennis and Schnabel, Numerical Methods for Unconstrained Optimization and Nonlinear Equations 1983, Prentice-Hall
`void`	`SetMaxFuncEval(int nMaxFuncEval)`	set limit for the number of function evaluations
`void`	`SetMaxIterations(int nMaxIterations)`	set limit for the number of iterations
`void`	`SetMethodPar(CDVector vMethodPar)`	set numerical gradient function
`int`	`SoftLineSearch(CDVector xc, double fc, CDVector g, CDVector sn, CDVector xplus, double[] fplus, boolean[] maxtaken)`	Perform line search Given g'p < 0 and alpha < 1/2 (alpha = 1e-4 is used), find plus = xc + lambda p,lambda in [0;1], such that f(xplus) <= f(xc) + alpha * lambda * g'p, using backtracking line search Algorithm A6.3.1 p. 325 Dennis and Schnabel, Numerical Methods for Unconstrained Optimization and Nonlinear Equations 1983, Prentice-Hall Input: xc: parameter fc: function value at xc g: gradient at xc sn: search direction Output: xplus: new parameter fplus: function value for new parameter maxtaken: max taken in line search return termination code
`int`	`UmStop(CDVector x, CDVector xplus, double f, double fplus, CDVector g, int retcode, boolean maxtaken)`	Decide wether to terminate minimization Modified version of Algorithm A7.2.1 p. 347 Dennis and Schnabel, Numerical Methods for Unconstrained Optimization and Nonlinear Equations 1983, Prentice-Hall Input: x: parameter xplus: new parameter fplus: function value for new parameter g: gradient at x retcode: return code from line search maxtaken: max taken in line search Output: return termination code
`int`	`UmStop0(CDVector x0, double functionValue, CDVector gradient)`

Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Field Detail
  - ENumGrad
```
public int ENumGrad
```
  - ForwardDifference
```
public static int ForwardDifference
```
  - CentralDifference
```
public static int CentralDifference
```
  - FitLine
```
public static int FitLine
```
  - EOptMethod
```
public int EOptMethod
```
  - eoptUnknown
```
public static int eoptUnknown
```
  - eoptSteepestDescent
```
public static int eoptSteepestDescent
```
  - eoptConjugateGradient
```
public static int eoptConjugateGradient
```
  - eoptBFGS
```
public static int eoptBFGS
```
  - eoptSimulatedAnnealing
```
public static int eoptSimulatedAnnealing
```
  - eoptPatternSearch
```
public static int eoptPatternSearch
```
  - ETermCode
```
public int ETermCode
```
  - etermNoStop
```
public static int etermNoStop
```
  - etermGradTol
```
public static int etermGradTol
```
  - etermStepTol
```
public static int etermStepTol
```
  - etermLineSearch
```
public static int etermLineSearch
```
  - etermConsecMaxStepMax
```
public static int etermConsecMaxStepMax
```
  - etermDeltaFuncVal
```
public static int etermDeltaFuncVal
```
  - etermMaxFuncEval
```
public static int etermMaxFuncEval
```
  - etermMaxIterations
```
public static int etermMaxIterations
```
  - etermUnknown
```
public static int etermUnknown
```
  - m_dMachEps
```
public double m_dMachEps
```
    Machine precision
  - m_dTypF
```
public double m_dTypF
```
    positive scalar estimating the magnitude of f(x) near he minimizer x-star. It is only used in the gradient stopping condition given below. typf should be approximately |f(x*)|
  - m_nFDigits
```
public int m_nFDigits
```
    A positive integer specifying the number of reliable digits returned by the objective function FN. fdigits is used to set the parameter n (eta) that is used in the code to specify the relative noise in f(x); the main use of eta is in calculation finite difference step size. eta is set to macheps if fdigits = -1. If f(x) is suspected to be noisy but the approximate value of fdigits is unknown, it should be estimated be the routine of Hamming[1973] given in Gill, Murray and Wright[1981]
  - m_dMaxStep
```
public double m_dMaxStep
```
    A positive scalar giving the maximum allowable scaled steplength at any iteration. maxstep is used to prevent steps that would cause the optimazation algorithm to overflow or leave the domain of interest, as well as to detect divergence. It should be chosen small enough to prevent the first two of these occurrences but larger than any anticipated reasonable stepsize. The algorithm will halt if it takes steps of length maxstep on m_nConsecMaxStepMax conseccutive iterations
  - m_nConsecMax
```
public int m_nConsecMax
```
    Number of conseccutive past steps whose scaled length was equal to maxstep
  - m_dEta
```
public double m_dEta
```
    machine precision.
  - m_fAnalyticGrad
```
public boolean m_fAnalyticGrad
```
    flag indicating if the gradient is to be calc. analyticaly or nummericaly
  - m_eNumGrad
```
public int m_eNumGrad
```
    methods for estimating the gradient nummericaly
  - m_iStopCriteria
```
public int m_iStopCriteria
```
    holds the stop criteria in use, e.g m_iStopCriteria = etermMaxFuncEval | etermMaxIterations
  - m_dGradTol
```
public double m_dGradTol
```
    A positive scalar giving the tolerance at which the scaled gradient in considered close enough to zero to terminate the algorithm
  - m_dStepTol
```
public double m_dStepTol
```
    A positive scalar giving the tolerance at which the scaled distance between two successive iterated is considered close enough to zero to terminate the algorithm
  - m_nConsecMaxStepMax
```
public int m_nConsecMaxStepMax
```
    max number of conseccutive past steps whose scaled length was equal to maxstep to terminate
  - m_dDeltaFuncVal
```
public double m_dDeltaFuncVal
```
    value for stop criterion: abs(f - fplus) < m_dDeltaFuncVal
  - m_nIterations
```
public int m_nIterations
```
    counter to the number of iterations.
  - m_nFuncEval
```
public int m_nFuncEval
```
    counter to the number of function evaluations.
  - m_nGradEval
```
public int m_nGradEval
```
    counter to the number of gradient evaluations.
  - m_fLogFuncValues
```
public boolean m_fLogFuncValues
```
    Logical variable which determines whether function parameters and corresponding return values should be stored for later analysis.
  - m_vFuncVal
```
public CDVector m_vFuncVal
```
  - m_vNFuncEval
```
public CDVector m_vNFuncEval
```
  - m_vvFuncParm
```
public java.util.Vector<CDVector> m_vvFuncParm
```
  - m_pFuncEvalBase
```
private CDOptimizeFuncBase m_pFuncEvalBase
```
    pointer to the function to be minimized.
  - m_vMethodPar
```
private CDVector m_vMethodPar
```
    special parameters used by each optimization method e.g. step size used for calc. nummerical gradient
  - m_nMaxIterations
```
private int m_nMaxIterations
```
    A positive integer specifying the maximum number of iterations that may be performed before the algorithm is halted. Appropriate values depend strongly on the dimension and difficulty of the problem, and the cost of evaluating the nonlinear function.
  - m_nMaxFuncEval
```
private int m_nMaxFuncEval
```
    limit for number of function evaluations.
- Constructor Detail
  - CDOptimizeBase
```
public CDOptimizeBase()
```
    Constructor
- Method Detail
  - Minimize
```
public abstract int Minimize(CDVector x,
                             CDOptimizeFuncBase pFuncEvalBase)
```
    the Minimize function using analytic gradient
    
    Parameters:
    
    x -
    
    pFuncEvalBase -
    
    Returns:
  - Name
```
public java.lang.String Name()
```
    name of optimization methode
    
    Returns:
  - OptMethod
```
public int OptMethod()
```
    name of optimization method
  - EvalFunction
```
public double EvalFunction(CDVector x)
```
    function and gradient evaluation methods
    
    Parameters:
    
    x -
    
    Returns:
  - EvalGradient
```
public void EvalGradient(CDVector x,
                         CDVector gc,
                         double dFuncVal)
```
    function and gradient evaluation methods
    
    Parameters:
    
    x -
    
    gc -
    
    dFuncVal -
  - MethodPar
```
public CDVector MethodPar()
```
    get numerical gradient function
    
    Returns:
  - SetMethodPar
```
public void SetMethodPar(CDVector vMethodPar)
```
    set numerical gradient function
    
    Parameters:
    
    vMethodPar -
  - MaxIterations
```
public int MaxIterations()
```
    get limit for the number of iterations
    
    Returns:
  - MaxFuncEval
```
public int MaxFuncEval()
```
    get the number of evaluations
    
    Returns:
  - SetFuncEvalBase
```
public void SetFuncEvalBase(CDOptimizeFuncBase pFuncEval)
```
    set point to func eval. base.
    
    Parameters:
    
    pFuncEval -
  - GetFuncEvalBase
```
public CDOptimizeFuncBase GetFuncEvalBase()
```
    get point to func eval. base.
    
    Returns:
  - SetMaxIterations
```
public void SetMaxIterations(int nMaxIterations)
```
    set limit for the number of iterations
    
    Parameters:
    
    nMaxIterations - max iterations
  - SetMaxFuncEval
```
public void SetMaxFuncEval(int nMaxFuncEval)
```
    set limit for the number of function evaluations
    
    Parameters:
    
    nMaxFuncEval - max number of evaluations.
  - SetMachineEps
```
public void SetMachineEps()
```
    ----------------------------[ MachineEps ]---------------------------- Calculate machine epsilon Algorithm A1.3.1 - p. 303 Dennis and Schnabel, Numerical Methods for Unconstrained Optimization and Nonlinear Equations 1983, Prentice-Hall
  - UmStop0
```
public int UmStop0(CDVector x0,
                   double functionValue,
                   CDVector gradient)
```
  - UmStop
```
public int UmStop(CDVector x,
                  CDVector xplus,
                  double f,
                  double fplus,
                  CDVector g,
                  int retcode,
                  boolean maxtaken)
```
    Decide wether to terminate minimization Modified version of Algorithm A7.2.1 p. 347 Dennis and Schnabel, Numerical Methods for Unconstrained Optimization and Nonlinear Equations 1983, Prentice-Hall Input: x: parameter xplus: new parameter fplus: function value for new parameter g: gradient at x retcode: return code from line search maxtaken: max taken in line search Output: return termination code
  - LineSearch
```
public int LineSearch(CDVector xc,
                      double fc,
                      CDVector g,
                      CDVector p,
                      CDVector xplus,
                      double[] fplus,
                      boolean[] maxtaken)
```
    Warpper for line search
    
    Parameters:
    
    xc - parameter
    
    fc - function value at xc
    
    g - gradient at xc
    
    p - search direction
    
    xplus - new parameter
    
    fplus - function value for new parameter
    
    maxtaken - max taken in line search
    
    Returns:
    
    termination code
  - LineSearch
```
public int LineSearch(CDVector xc,
                      double fc,
                      CDVector g,
                      CDVector p,
                      CDVector xplus,
                      double[] fplus,
                      boolean[] maxtaken,
                      boolean fSoft)
```
  - ExactLineSearch
```
public int ExactLineSearch(CDVector xc,
                           double fc,
                           CDVector g,
                           CDVector p,
                           CDVector xplus,
                           double[] fplus,
                           boolean[] maxtaken)
```
    Perform exact line search Input: xc: parameter fc: function value at xc g: gradient at xc p: search direction Output: xplus: new parameter fplus: function value for new parameter maxtaken: max taken in line search return termination code
  - SoftLineSearch
```
public int SoftLineSearch(CDVector xc,
                          double fc,
                          CDVector g,
                          CDVector sn,
                          CDVector xplus,
                          double[] fplus,
                          boolean[] maxtaken)
```
    Perform line search Given g'p < 0 and alpha < 1/2 (alpha = 1e-4 is used), find plus = xc + lambda p,lambda in [0;1], such that f(xplus) <= f(xc) + alpha * lambda * g'p, using backtracking line search Algorithm A6.3.1 p. 325 Dennis and Schnabel, Numerical Methods for Unconstrained Optimization and Nonlinear Equations 1983, Prentice-Hall Input: xc: parameter fc: function value at xc g: gradient at xc sn: search direction Output: xplus: new parameter fplus: function value for new parameter maxtaken: max taken in line search return termination code
  - NumGrad
```
public void NumGrad(CDVector x,
                    CDVector gradient,
                    double dFuncVal)
```
    calculate nummerical gradient function usingdecided gradient calculation method input: x: parameter dFuncVal: function value in x output: gradient: gradient
  - MinimizeNum
```
public int MinimizeNum(CDVector x,
                       CDOptimizeFuncBase pFuncEvalBase,
                       CDVector vMethodPar)
```

Class CDOptimizeBase

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

ENumGrad

ForwardDifference

CentralDifference

FitLine

EOptMethod

eoptUnknown

eoptSteepestDescent

eoptConjugateGradient

eoptBFGS

eoptSimulatedAnnealing

eoptPatternSearch

ETermCode

etermNoStop

etermGradTol

etermStepTol

etermLineSearch

etermConsecMaxStepMax

etermDeltaFuncVal

etermMaxFuncEval

etermMaxIterations

etermUnknown

m_dMachEps

m_dTypF

m_nFDigits

m_dMaxStep

m_nConsecMax

m_dEta

m_fAnalyticGrad

m_eNumGrad

m_iStopCriteria

m_dGradTol

m_dStepTol

m_nConsecMaxStepMax

m_dDeltaFuncVal

m_nIterations

m_nFuncEval

m_nGradEval

m_fLogFuncValues

m_vFuncVal

m_vNFuncEval

m_vvFuncParm

m_pFuncEvalBase

m_vMethodPar

m_nMaxIterations

m_nMaxFuncEval

Constructor Detail

CDOptimizeBase

Method Detail

Minimize

Name

OptMethod

EvalFunction

EvalGradient

MethodPar

SetMethodPar

MaxIterations

MaxFuncEval

SetFuncEvalBase

GetFuncEvalBase

SetMaxIterations

SetMaxFuncEval

SetMachineEps

UmStop0

UmStop

LineSearch

LineSearch

ExactLineSearch

SoftLineSearch

NumGrad

MinimizeNum